[ic] Should I buy another CPU? - rpc mode - MaxServers 0

Tue Mar 19 05:42:01 2002

From: "Kyle Cook" <kyle@invisio.com>
To: <interchange-users@interchange.redhat.com>
Sent: Tuesday, 19 March 2002 2:00
Subject: RE: [ic] Should I buy another CPU?

> At 05:35 PM 3/18/02, you wrote:
> > > From: interchange-users-admin@interchange.redhat.com
> >[mailto:interchange-
> > > users-admin@interchange.redhat.com] On Behalf Of John Beima
> > >
> ><-- SNIP -->
> > > One last thing you need to look at is drive speed and choices.... If
> >you
> > > are
> > > experencing on and off failures in pgp encryption and/or payment
> >gateways.
> > > This
> > > is more likely than not, your drives are too slow. These
> >programs/modules
> > > need
> > > to make tmp files, they also need RAM, and if they can't get both fast
> > > enough,
> > > they fail. You will see this problem a lot on the list, if people
> >actually
> > > READ
> > > past messsages, they would see it and STOP posting it again and again.
> >
> >This is obviously directed at me and others who experienced this
> >problem.  Mike Heins and Orko from the list spent a considerable amount
> >of time working with me on this issue and we found that it was not due
> >to RAM or HD speed.  I posted before on this issue that we run U2W SCSI
> >drives and have more then enough free ram for the one site that is being
> >served.
> >
> >The problem being experienced is due to a bug in Perl's signal handling
> >and can be fixed by using the applying the fix that Mike came up with
> >(RPC mode, MaxServers 0).  If we would have stopped posting this problem
> >to the list the first time it would have never been fixed.  I'm not sure
> >why you are upset that this problem was posted to the list, but due to
> >Mike wanting to get to the bottom of this and more then one person
> >experiencing the problem a fix was found... and it was not to simply
> >throw more hardware at the problem.
> >
> >-Ron
>
> Occasional problems could be the result of insufficient resources
> or the "perl signal handling bug".  If you don't see resource errors
> in the system log files (not counting ic log files) then try rpc mode
> with MaxServers set to 0.
>
> For me this has solved the problem.
>
>
> Just to clarify on settings to change IF you experience occasional
> failures of external programs (sendmail,pgp,LockOut Command, etc)
>
> interchange.cfg :
>
> Variable  TRAFFIC  rpc
>
>
> ifdef TRAFFIC =~ /rpc/i
> Message RPC traffic settings.
> PreFork             Yes
> StartServers        5
> MaxRequestsPerChild 25
> HouseKeeping        2
> PIDcheck            120
> MaxServers          0
> endif

Terrific, so far so good! Have tested a dozen orders and where previously
only 25% included encrypted information so far the success rate is 100%

Thanks for posting this as it means I may be able to live without updating
the current hardware for a time!